"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630160919.4761047/events.out.tfevents.1630160919.86bb0ddabf9b.4092.191 +3 -0
- model-bin/finetune/base/log/1630161303.291062/events.out.tfevents.1630161303.86bb0ddabf9b.4092.193 +3 -0
- model-bin/finetune/base/log/1630161692.6533394/events.out.tfevents.1630161692.86bb0ddabf9b.4092.195 +3 -0
- model-bin/finetune/base/log/1630162085.8766842/events.out.tfevents.1630162085.86bb0ddabf9b.4092.197 +3 -0
- model-bin/finetune/base/log/1630162481.5400343/events.out.tfevents.1630162481.86bb0ddabf9b.4092.199 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630160919.86bb0ddabf9b.4092.190 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630161303.86bb0ddabf9b.4092.192 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630161692.86bb0ddabf9b.4092.194 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630162085.86bb0ddabf9b.4092.196 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630162481.86bb0ddabf9b.4092.198 +3 -0
model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165393
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b1c26c07fe7d9e76f7c1a7c2d338623525476591507c330a369916832c4b6ba0
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/pytorch_model.bin
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 377909911
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1920d5bcd50979c20fb7231b1a06e8dd49cbcc0e3955c4f8d624d1201f658d3d
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0862150e25fe101b150152dabd167c392fca4124550c0d8ab85931d40a8ebf6
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/scaler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 559
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dff8ba8ac37ba987a6e86e03d5f7a18a2e9fd17337acbc791d9bb04621ee7d3a
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/scheduler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 623
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4227d8f89a196fb496e38569a6699f58517114ed4b07d9cb88e1b2b7038d30c6
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17105075053609722,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-129674",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -264507,11 +264507,806 @@
|
|
| 264507 |
"eval_steps_per_second": 0.794,
|
| 264508 |
"eval_wer": 0.18174261316405954,
|
| 264509 |
"step": 129799
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 264510 |
}
|
| 264511 |
],
|
| 264512 |
-
"max_steps":
|
| 264513 |
"num_train_epochs": 5000,
|
| 264514 |
-
"total_flos": 3.
|
| 264515 |
"trial_name": null,
|
| 264516 |
"trial_params": null
|
| 264517 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17105075053609722,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-129674",
|
| 4 |
+
"epoch": 1050.995983935743,
|
| 5 |
+
"global_step": 130420,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 264507 |
"eval_steps_per_second": 0.794,
|
| 264508 |
"eval_wer": 0.18174261316405954,
|
| 264509 |
"step": 129799
|
| 264510 |
+
},
|
| 264511 |
+
{
|
| 264512 |
+
"epoch": 1046.01,
|
| 264513 |
+
"learning_rate": 7.936714743589745e-06,
|
| 264514 |
+
"loss": 0.4129,
|
| 264515 |
+
"step": 129800
|
| 264516 |
+
},
|
| 264517 |
+
{
|
| 264518 |
+
"epoch": 1046.05,
|
| 264519 |
+
"learning_rate": 7.936634615384616e-06,
|
| 264520 |
+
"loss": 0.2893,
|
| 264521 |
+
"step": 129805
|
| 264522 |
+
},
|
| 264523 |
+
{
|
| 264524 |
+
"epoch": 1046.09,
|
| 264525 |
+
"learning_rate": 7.936554487179487e-06,
|
| 264526 |
+
"loss": 0.2299,
|
| 264527 |
+
"step": 129810
|
| 264528 |
+
},
|
| 264529 |
+
{
|
| 264530 |
+
"epoch": 1046.13,
|
| 264531 |
+
"learning_rate": 7.93647435897436e-06,
|
| 264532 |
+
"loss": 0.2756,
|
| 264533 |
+
"step": 129815
|
| 264534 |
+
},
|
| 264535 |
+
{
|
| 264536 |
+
"epoch": 1046.17,
|
| 264537 |
+
"learning_rate": 7.936394230769232e-06,
|
| 264538 |
+
"loss": 0.4417,
|
| 264539 |
+
"step": 129820
|
| 264540 |
+
},
|
| 264541 |
+
{
|
| 264542 |
+
"epoch": 1046.21,
|
| 264543 |
+
"learning_rate": 7.936314102564103e-06,
|
| 264544 |
+
"loss": 1.0294,
|
| 264545 |
+
"step": 129825
|
| 264546 |
+
},
|
| 264547 |
+
{
|
| 264548 |
+
"epoch": 1046.25,
|
| 264549 |
+
"learning_rate": 7.936233974358974e-06,
|
| 264550 |
+
"loss": 0.2665,
|
| 264551 |
+
"step": 129830
|
| 264552 |
+
},
|
| 264553 |
+
{
|
| 264554 |
+
"epoch": 1046.29,
|
| 264555 |
+
"learning_rate": 7.936153846153847e-06,
|
| 264556 |
+
"loss": 0.3326,
|
| 264557 |
+
"step": 129835
|
| 264558 |
+
},
|
| 264559 |
+
{
|
| 264560 |
+
"epoch": 1046.33,
|
| 264561 |
+
"learning_rate": 7.936073717948719e-06,
|
| 264562 |
+
"loss": 0.3502,
|
| 264563 |
+
"step": 129840
|
| 264564 |
+
},
|
| 264565 |
+
{
|
| 264566 |
+
"epoch": 1046.37,
|
| 264567 |
+
"learning_rate": 7.93599358974359e-06,
|
| 264568 |
+
"loss": 0.5617,
|
| 264569 |
+
"step": 129845
|
| 264570 |
+
},
|
| 264571 |
+
{
|
| 264572 |
+
"epoch": 1046.41,
|
| 264573 |
+
"learning_rate": 7.935913461538463e-06,
|
| 264574 |
+
"loss": 1.1194,
|
| 264575 |
+
"step": 129850
|
| 264576 |
+
},
|
| 264577 |
+
{
|
| 264578 |
+
"epoch": 1046.45,
|
| 264579 |
+
"learning_rate": 7.935833333333333e-06,
|
| 264580 |
+
"loss": 0.3628,
|
| 264581 |
+
"step": 129855
|
| 264582 |
+
},
|
| 264583 |
+
{
|
| 264584 |
+
"epoch": 1046.49,
|
| 264585 |
+
"learning_rate": 7.935753205128206e-06,
|
| 264586 |
+
"loss": 0.2528,
|
| 264587 |
+
"step": 129860
|
| 264588 |
+
},
|
| 264589 |
+
{
|
| 264590 |
+
"epoch": 1046.53,
|
| 264591 |
+
"learning_rate": 7.935673076923077e-06,
|
| 264592 |
+
"loss": 0.3784,
|
| 264593 |
+
"step": 129865
|
| 264594 |
+
},
|
| 264595 |
+
{
|
| 264596 |
+
"epoch": 1046.57,
|
| 264597 |
+
"learning_rate": 7.935592948717949e-06,
|
| 264598 |
+
"loss": 0.478,
|
| 264599 |
+
"step": 129870
|
| 264600 |
+
},
|
| 264601 |
+
{
|
| 264602 |
+
"epoch": 1046.61,
|
| 264603 |
+
"learning_rate": 7.93551282051282e-06,
|
| 264604 |
+
"loss": 1.186,
|
| 264605 |
+
"step": 129875
|
| 264606 |
+
},
|
| 264607 |
+
{
|
| 264608 |
+
"epoch": 1046.65,
|
| 264609 |
+
"learning_rate": 7.935432692307693e-06,
|
| 264610 |
+
"loss": 0.3365,
|
| 264611 |
+
"step": 129880
|
| 264612 |
+
},
|
| 264613 |
+
{
|
| 264614 |
+
"epoch": 1046.69,
|
| 264615 |
+
"learning_rate": 7.935352564102566e-06,
|
| 264616 |
+
"loss": 0.3178,
|
| 264617 |
+
"step": 129885
|
| 264618 |
+
},
|
| 264619 |
+
{
|
| 264620 |
+
"epoch": 1046.73,
|
| 264621 |
+
"learning_rate": 7.935272435897436e-06,
|
| 264622 |
+
"loss": 0.4071,
|
| 264623 |
+
"step": 129890
|
| 264624 |
+
},
|
| 264625 |
+
{
|
| 264626 |
+
"epoch": 1046.77,
|
| 264627 |
+
"learning_rate": 7.935192307692309e-06,
|
| 264628 |
+
"loss": 0.4907,
|
| 264629 |
+
"step": 129895
|
| 264630 |
+
},
|
| 264631 |
+
{
|
| 264632 |
+
"epoch": 1046.81,
|
| 264633 |
+
"learning_rate": 7.93511217948718e-06,
|
| 264634 |
+
"loss": 0.9865,
|
| 264635 |
+
"step": 129900
|
| 264636 |
+
},
|
| 264637 |
+
{
|
| 264638 |
+
"epoch": 1046.85,
|
| 264639 |
+
"learning_rate": 7.935032051282052e-06,
|
| 264640 |
+
"loss": 0.2768,
|
| 264641 |
+
"step": 129905
|
| 264642 |
+
},
|
| 264643 |
+
{
|
| 264644 |
+
"epoch": 1046.89,
|
| 264645 |
+
"learning_rate": 7.934951923076923e-06,
|
| 264646 |
+
"loss": 0.2493,
|
| 264647 |
+
"step": 129910
|
| 264648 |
+
},
|
| 264649 |
+
{
|
| 264650 |
+
"epoch": 1046.93,
|
| 264651 |
+
"learning_rate": 7.934871794871796e-06,
|
| 264652 |
+
"loss": 0.2988,
|
| 264653 |
+
"step": 129915
|
| 264654 |
+
},
|
| 264655 |
+
{
|
| 264656 |
+
"epoch": 1046.97,
|
| 264657 |
+
"learning_rate": 7.934791666666667e-06,
|
| 264658 |
+
"loss": 0.5019,
|
| 264659 |
+
"step": 129920
|
| 264660 |
+
},
|
| 264661 |
+
{
|
| 264662 |
+
"epoch": 1047.0,
|
| 264663 |
+
"eval_loss": 0.38806846737861633,
|
| 264664 |
+
"eval_runtime": 35.8177,
|
| 264665 |
+
"eval_samples_per_second": 23.313,
|
| 264666 |
+
"eval_steps_per_second": 0.754,
|
| 264667 |
+
"eval_wer": 0.17985185185185185,
|
| 264668 |
+
"step": 129923
|
| 264669 |
+
},
|
| 264670 |
+
{
|
| 264671 |
+
"epoch": 1047.02,
|
| 264672 |
+
"learning_rate": 7.934711538461539e-06,
|
| 264673 |
+
"loss": 0.3428,
|
| 264674 |
+
"step": 129925
|
| 264675 |
+
},
|
| 264676 |
+
{
|
| 264677 |
+
"epoch": 1047.06,
|
| 264678 |
+
"learning_rate": 7.93463141025641e-06,
|
| 264679 |
+
"loss": 0.3148,
|
| 264680 |
+
"step": 129930
|
| 264681 |
+
},
|
| 264682 |
+
{
|
| 264683 |
+
"epoch": 1047.1,
|
| 264684 |
+
"learning_rate": 7.934551282051283e-06,
|
| 264685 |
+
"loss": 0.341,
|
| 264686 |
+
"step": 129935
|
| 264687 |
+
},
|
| 264688 |
+
{
|
| 264689 |
+
"epoch": 1047.14,
|
| 264690 |
+
"learning_rate": 7.934471153846154e-06,
|
| 264691 |
+
"loss": 0.2991,
|
| 264692 |
+
"step": 129940
|
| 264693 |
+
},
|
| 264694 |
+
{
|
| 264695 |
+
"epoch": 1047.18,
|
| 264696 |
+
"learning_rate": 7.934391025641026e-06,
|
| 264697 |
+
"loss": 0.7212,
|
| 264698 |
+
"step": 129945
|
| 264699 |
+
},
|
| 264700 |
+
{
|
| 264701 |
+
"epoch": 1047.22,
|
| 264702 |
+
"learning_rate": 7.934310897435899e-06,
|
| 264703 |
+
"loss": 1.0056,
|
| 264704 |
+
"step": 129950
|
| 264705 |
+
},
|
| 264706 |
+
{
|
| 264707 |
+
"epoch": 1047.26,
|
| 264708 |
+
"learning_rate": 7.93423076923077e-06,
|
| 264709 |
+
"loss": 0.2669,
|
| 264710 |
+
"step": 129955
|
| 264711 |
+
},
|
| 264712 |
+
{
|
| 264713 |
+
"epoch": 1047.3,
|
| 264714 |
+
"learning_rate": 7.934150641025642e-06,
|
| 264715 |
+
"loss": 0.2889,
|
| 264716 |
+
"step": 129960
|
| 264717 |
+
},
|
| 264718 |
+
{
|
| 264719 |
+
"epoch": 1047.34,
|
| 264720 |
+
"learning_rate": 7.934070512820513e-06,
|
| 264721 |
+
"loss": 0.3207,
|
| 264722 |
+
"step": 129965
|
| 264723 |
+
},
|
| 264724 |
+
{
|
| 264725 |
+
"epoch": 1047.38,
|
| 264726 |
+
"learning_rate": 7.933990384615386e-06,
|
| 264727 |
+
"loss": 0.7011,
|
| 264728 |
+
"step": 129970
|
| 264729 |
+
},
|
| 264730 |
+
{
|
| 264731 |
+
"epoch": 1047.42,
|
| 264732 |
+
"learning_rate": 7.933910256410257e-06,
|
| 264733 |
+
"loss": 0.9856,
|
| 264734 |
+
"step": 129975
|
| 264735 |
+
},
|
| 264736 |
+
{
|
| 264737 |
+
"epoch": 1047.46,
|
| 264738 |
+
"learning_rate": 7.933830128205129e-06,
|
| 264739 |
+
"loss": 0.2456,
|
| 264740 |
+
"step": 129980
|
| 264741 |
+
},
|
| 264742 |
+
{
|
| 264743 |
+
"epoch": 1047.5,
|
| 264744 |
+
"learning_rate": 7.933750000000002e-06,
|
| 264745 |
+
"loss": 0.2615,
|
| 264746 |
+
"step": 129985
|
| 264747 |
+
},
|
| 264748 |
+
{
|
| 264749 |
+
"epoch": 1047.54,
|
| 264750 |
+
"learning_rate": 7.933669871794873e-06,
|
| 264751 |
+
"loss": 0.4054,
|
| 264752 |
+
"step": 129990
|
| 264753 |
+
},
|
| 264754 |
+
{
|
| 264755 |
+
"epoch": 1047.58,
|
| 264756 |
+
"learning_rate": 7.933589743589744e-06,
|
| 264757 |
+
"loss": 0.5305,
|
| 264758 |
+
"step": 129995
|
| 264759 |
+
},
|
| 264760 |
+
{
|
| 264761 |
+
"epoch": 1047.62,
|
| 264762 |
+
"learning_rate": 7.933509615384616e-06,
|
| 264763 |
+
"loss": 0.8806,
|
| 264764 |
+
"step": 130000
|
| 264765 |
+
},
|
| 264766 |
+
{
|
| 264767 |
+
"epoch": 1047.66,
|
| 264768 |
+
"learning_rate": 7.933429487179489e-06,
|
| 264769 |
+
"loss": 0.2519,
|
| 264770 |
+
"step": 130005
|
| 264771 |
+
},
|
| 264772 |
+
{
|
| 264773 |
+
"epoch": 1047.7,
|
| 264774 |
+
"learning_rate": 7.933349358974359e-06,
|
| 264775 |
+
"loss": 0.2915,
|
| 264776 |
+
"step": 130010
|
| 264777 |
+
},
|
| 264778 |
+
{
|
| 264779 |
+
"epoch": 1047.74,
|
| 264780 |
+
"learning_rate": 7.933269230769232e-06,
|
| 264781 |
+
"loss": 0.2979,
|
| 264782 |
+
"step": 130015
|
| 264783 |
+
},
|
| 264784 |
+
{
|
| 264785 |
+
"epoch": 1047.78,
|
| 264786 |
+
"learning_rate": 7.933189102564103e-06,
|
| 264787 |
+
"loss": 0.569,
|
| 264788 |
+
"step": 130020
|
| 264789 |
+
},
|
| 264790 |
+
{
|
| 264791 |
+
"epoch": 1047.82,
|
| 264792 |
+
"learning_rate": 7.933108974358974e-06,
|
| 264793 |
+
"loss": 0.9575,
|
| 264794 |
+
"step": 130025
|
| 264795 |
+
},
|
| 264796 |
+
{
|
| 264797 |
+
"epoch": 1047.86,
|
| 264798 |
+
"learning_rate": 7.933028846153846e-06,
|
| 264799 |
+
"loss": 0.281,
|
| 264800 |
+
"step": 130030
|
| 264801 |
+
},
|
| 264802 |
+
{
|
| 264803 |
+
"epoch": 1047.9,
|
| 264804 |
+
"learning_rate": 7.932948717948719e-06,
|
| 264805 |
+
"loss": 0.2734,
|
| 264806 |
+
"step": 130035
|
| 264807 |
+
},
|
| 264808 |
+
{
|
| 264809 |
+
"epoch": 1047.94,
|
| 264810 |
+
"learning_rate": 7.93286858974359e-06,
|
| 264811 |
+
"loss": 0.3178,
|
| 264812 |
+
"step": 130040
|
| 264813 |
+
},
|
| 264814 |
+
{
|
| 264815 |
+
"epoch": 1047.98,
|
| 264816 |
+
"learning_rate": 7.932788461538461e-06,
|
| 264817 |
+
"loss": 0.768,
|
| 264818 |
+
"step": 130045
|
| 264819 |
+
},
|
| 264820 |
+
{
|
| 264821 |
+
"epoch": 1048.0,
|
| 264822 |
+
"eval_loss": 0.3541869521141052,
|
| 264823 |
+
"eval_runtime": 34.4713,
|
| 264824 |
+
"eval_samples_per_second": 24.223,
|
| 264825 |
+
"eval_steps_per_second": 0.783,
|
| 264826 |
+
"eval_wer": 0.17978857100613588,
|
| 264827 |
+
"step": 130047
|
| 264828 |
+
},
|
| 264829 |
+
{
|
| 264830 |
+
"epoch": 1048.02,
|
| 264831 |
+
"learning_rate": 7.932708333333334e-06,
|
| 264832 |
+
"loss": 0.3823,
|
| 264833 |
+
"step": 130050
|
| 264834 |
+
},
|
| 264835 |
+
{
|
| 264836 |
+
"epoch": 1048.06,
|
| 264837 |
+
"learning_rate": 7.932628205128206e-06,
|
| 264838 |
+
"loss": 0.2531,
|
| 264839 |
+
"step": 130055
|
| 264840 |
+
},
|
| 264841 |
+
{
|
| 264842 |
+
"epoch": 1048.1,
|
| 264843 |
+
"learning_rate": 7.932548076923077e-06,
|
| 264844 |
+
"loss": 0.3504,
|
| 264845 |
+
"step": 130060
|
| 264846 |
+
},
|
| 264847 |
+
{
|
| 264848 |
+
"epoch": 1048.14,
|
| 264849 |
+
"learning_rate": 7.932467948717949e-06,
|
| 264850 |
+
"loss": 0.3598,
|
| 264851 |
+
"step": 130065
|
| 264852 |
+
},
|
| 264853 |
+
{
|
| 264854 |
+
"epoch": 1048.18,
|
| 264855 |
+
"learning_rate": 7.932387820512822e-06,
|
| 264856 |
+
"loss": 0.7025,
|
| 264857 |
+
"step": 130070
|
| 264858 |
+
},
|
| 264859 |
+
{
|
| 264860 |
+
"epoch": 1048.22,
|
| 264861 |
+
"learning_rate": 7.932307692307693e-06,
|
| 264862 |
+
"loss": 0.8204,
|
| 264863 |
+
"step": 130075
|
| 264864 |
+
},
|
| 264865 |
+
{
|
| 264866 |
+
"epoch": 1048.27,
|
| 264867 |
+
"learning_rate": 7.932227564102564e-06,
|
| 264868 |
+
"loss": 0.3144,
|
| 264869 |
+
"step": 130080
|
| 264870 |
+
},
|
| 264871 |
+
{
|
| 264872 |
+
"epoch": 1048.31,
|
| 264873 |
+
"learning_rate": 7.932147435897437e-06,
|
| 264874 |
+
"loss": 0.2883,
|
| 264875 |
+
"step": 130085
|
| 264876 |
+
},
|
| 264877 |
+
{
|
| 264878 |
+
"epoch": 1048.35,
|
| 264879 |
+
"learning_rate": 7.932067307692309e-06,
|
| 264880 |
+
"loss": 0.3835,
|
| 264881 |
+
"step": 130090
|
| 264882 |
+
},
|
| 264883 |
+
{
|
| 264884 |
+
"epoch": 1048.39,
|
| 264885 |
+
"learning_rate": 7.93198717948718e-06,
|
| 264886 |
+
"loss": 0.8257,
|
| 264887 |
+
"step": 130095
|
| 264888 |
+
},
|
| 264889 |
+
{
|
| 264890 |
+
"epoch": 1048.43,
|
| 264891 |
+
"learning_rate": 7.931907051282051e-06,
|
| 264892 |
+
"loss": 0.8124,
|
| 264893 |
+
"step": 130100
|
| 264894 |
+
},
|
| 264895 |
+
{
|
| 264896 |
+
"epoch": 1048.47,
|
| 264897 |
+
"learning_rate": 7.931826923076925e-06,
|
| 264898 |
+
"loss": 0.2594,
|
| 264899 |
+
"step": 130105
|
| 264900 |
+
},
|
| 264901 |
+
{
|
| 264902 |
+
"epoch": 1048.51,
|
| 264903 |
+
"learning_rate": 7.931746794871796e-06,
|
| 264904 |
+
"loss": 0.2822,
|
| 264905 |
+
"step": 130110
|
| 264906 |
+
},
|
| 264907 |
+
{
|
| 264908 |
+
"epoch": 1048.55,
|
| 264909 |
+
"learning_rate": 7.931666666666667e-06,
|
| 264910 |
+
"loss": 0.3836,
|
| 264911 |
+
"step": 130115
|
| 264912 |
+
},
|
| 264913 |
+
{
|
| 264914 |
+
"epoch": 1048.59,
|
| 264915 |
+
"learning_rate": 7.931586538461539e-06,
|
| 264916 |
+
"loss": 0.7467,
|
| 264917 |
+
"step": 130120
|
| 264918 |
+
},
|
| 264919 |
+
{
|
| 264920 |
+
"epoch": 1048.63,
|
| 264921 |
+
"learning_rate": 7.931506410256412e-06,
|
| 264922 |
+
"loss": 0.8817,
|
| 264923 |
+
"step": 130125
|
| 264924 |
+
},
|
| 264925 |
+
{
|
| 264926 |
+
"epoch": 1048.67,
|
| 264927 |
+
"learning_rate": 7.931426282051283e-06,
|
| 264928 |
+
"loss": 0.3068,
|
| 264929 |
+
"step": 130130
|
| 264930 |
+
},
|
| 264931 |
+
{
|
| 264932 |
+
"epoch": 1048.71,
|
| 264933 |
+
"learning_rate": 7.931346153846154e-06,
|
| 264934 |
+
"loss": 0.2516,
|
| 264935 |
+
"step": 130135
|
| 264936 |
+
},
|
| 264937 |
+
{
|
| 264938 |
+
"epoch": 1048.75,
|
| 264939 |
+
"learning_rate": 7.931266025641027e-06,
|
| 264940 |
+
"loss": 0.3328,
|
| 264941 |
+
"step": 130140
|
| 264942 |
+
},
|
| 264943 |
+
{
|
| 264944 |
+
"epoch": 1048.79,
|
| 264945 |
+
"learning_rate": 7.931185897435899e-06,
|
| 264946 |
+
"loss": 0.871,
|
| 264947 |
+
"step": 130145
|
| 264948 |
+
},
|
| 264949 |
+
{
|
| 264950 |
+
"epoch": 1048.83,
|
| 264951 |
+
"learning_rate": 7.93110576923077e-06,
|
| 264952 |
+
"loss": 0.9983,
|
| 264953 |
+
"step": 130150
|
| 264954 |
+
},
|
| 264955 |
+
{
|
| 264956 |
+
"epoch": 1048.87,
|
| 264957 |
+
"learning_rate": 7.931025641025641e-06,
|
| 264958 |
+
"loss": 0.2637,
|
| 264959 |
+
"step": 130155
|
| 264960 |
+
},
|
| 264961 |
+
{
|
| 264962 |
+
"epoch": 1048.91,
|
| 264963 |
+
"learning_rate": 7.930945512820515e-06,
|
| 264964 |
+
"loss": 0.3195,
|
| 264965 |
+
"step": 130160
|
| 264966 |
+
},
|
| 264967 |
+
{
|
| 264968 |
+
"epoch": 1048.95,
|
| 264969 |
+
"learning_rate": 7.930865384615384e-06,
|
| 264970 |
+
"loss": 0.3649,
|
| 264971 |
+
"step": 130165
|
| 264972 |
+
},
|
| 264973 |
+
{
|
| 264974 |
+
"epoch": 1048.99,
|
| 264975 |
+
"learning_rate": 7.930785256410257e-06,
|
| 264976 |
+
"loss": 0.7069,
|
| 264977 |
+
"step": 130170
|
| 264978 |
+
},
|
| 264979 |
+
{
|
| 264980 |
+
"epoch": 1049.0,
|
| 264981 |
+
"eval_loss": 0.4282522201538086,
|
| 264982 |
+
"eval_runtime": 37.3121,
|
| 264983 |
+
"eval_samples_per_second": 22.406,
|
| 264984 |
+
"eval_steps_per_second": 0.724,
|
| 264985 |
+
"eval_wer": 0.17979154603358424,
|
| 264986 |
+
"step": 130171
|
| 264987 |
+
},
|
| 264988 |
+
{
|
| 264989 |
+
"epoch": 1041.03,
|
| 264990 |
+
"learning_rate": 7.930705128205129e-06,
|
| 264991 |
+
"loss": 0.3971,
|
| 264992 |
+
"step": 130175
|
| 264993 |
+
},
|
| 264994 |
+
{
|
| 264995 |
+
"epoch": 1041.07,
|
| 264996 |
+
"learning_rate": 7.930625e-06,
|
| 264997 |
+
"loss": 0.2518,
|
| 264998 |
+
"step": 130180
|
| 264999 |
+
},
|
| 265000 |
+
{
|
| 265001 |
+
"epoch": 1041.11,
|
| 265002 |
+
"learning_rate": 7.930544871794873e-06,
|
| 265003 |
+
"loss": 0.3056,
|
| 265004 |
+
"step": 130185
|
| 265005 |
+
},
|
| 265006 |
+
{
|
| 265007 |
+
"epoch": 1041.15,
|
| 265008 |
+
"learning_rate": 7.930464743589744e-06,
|
| 265009 |
+
"loss": 0.3826,
|
| 265010 |
+
"step": 130190
|
| 265011 |
+
},
|
| 265012 |
+
{
|
| 265013 |
+
"epoch": 1041.19,
|
| 265014 |
+
"learning_rate": 7.930384615384616e-06,
|
| 265015 |
+
"loss": 0.9553,
|
| 265016 |
+
"step": 130195
|
| 265017 |
+
},
|
| 265018 |
+
{
|
| 265019 |
+
"epoch": 1041.23,
|
| 265020 |
+
"learning_rate": 7.930304487179487e-06,
|
| 265021 |
+
"loss": 0.6101,
|
| 265022 |
+
"step": 130200
|
| 265023 |
+
},
|
| 265024 |
+
{
|
| 265025 |
+
"epoch": 1041.27,
|
| 265026 |
+
"learning_rate": 7.93022435897436e-06,
|
| 265027 |
+
"loss": 0.2316,
|
| 265028 |
+
"step": 130205
|
| 265029 |
+
},
|
| 265030 |
+
{
|
| 265031 |
+
"epoch": 1041.31,
|
| 265032 |
+
"learning_rate": 7.930144230769232e-06,
|
| 265033 |
+
"loss": 0.3121,
|
| 265034 |
+
"step": 130210
|
| 265035 |
+
},
|
| 265036 |
+
{
|
| 265037 |
+
"epoch": 1041.35,
|
| 265038 |
+
"learning_rate": 7.930064102564103e-06,
|
| 265039 |
+
"loss": 0.4402,
|
| 265040 |
+
"step": 130215
|
| 265041 |
+
},
|
| 265042 |
+
{
|
| 265043 |
+
"epoch": 1041.39,
|
| 265044 |
+
"learning_rate": 7.929983974358974e-06,
|
| 265045 |
+
"loss": 0.9923,
|
| 265046 |
+
"step": 130220
|
| 265047 |
+
},
|
| 265048 |
+
{
|
| 265049 |
+
"epoch": 1041.43,
|
| 265050 |
+
"learning_rate": 7.929903846153847e-06,
|
| 265051 |
+
"loss": 0.607,
|
| 265052 |
+
"step": 130225
|
| 265053 |
+
},
|
| 265054 |
+
{
|
| 265055 |
+
"epoch": 1041.47,
|
| 265056 |
+
"learning_rate": 7.929823717948719e-06,
|
| 265057 |
+
"loss": 0.2848,
|
| 265058 |
+
"step": 130230
|
| 265059 |
+
},
|
| 265060 |
+
{
|
| 265061 |
+
"epoch": 1041.51,
|
| 265062 |
+
"learning_rate": 7.92974358974359e-06,
|
| 265063 |
+
"loss": 0.2927,
|
| 265064 |
+
"step": 130235
|
| 265065 |
+
},
|
| 265066 |
+
{
|
| 265067 |
+
"epoch": 1041.55,
|
| 265068 |
+
"learning_rate": 7.929663461538463e-06,
|
| 265069 |
+
"loss": 0.4192,
|
| 265070 |
+
"step": 130240
|
| 265071 |
+
},
|
| 265072 |
+
{
|
| 265073 |
+
"epoch": 1041.59,
|
| 265074 |
+
"learning_rate": 7.929583333333334e-06,
|
| 265075 |
+
"loss": 0.825,
|
| 265076 |
+
"step": 130245
|
| 265077 |
+
},
|
| 265078 |
+
{
|
| 265079 |
+
"epoch": 1041.63,
|
| 265080 |
+
"learning_rate": 7.929503205128206e-06,
|
| 265081 |
+
"loss": 0.736,
|
| 265082 |
+
"step": 130250
|
| 265083 |
+
},
|
| 265084 |
+
{
|
| 265085 |
+
"epoch": 1041.67,
|
| 265086 |
+
"learning_rate": 7.929423076923077e-06,
|
| 265087 |
+
"loss": 0.3488,
|
| 265088 |
+
"step": 130255
|
| 265089 |
+
},
|
| 265090 |
+
{
|
| 265091 |
+
"epoch": 1041.71,
|
| 265092 |
+
"learning_rate": 7.92934294871795e-06,
|
| 265093 |
+
"loss": 0.315,
|
| 265094 |
+
"step": 130260
|
| 265095 |
+
},
|
| 265096 |
+
{
|
| 265097 |
+
"epoch": 1041.75,
|
| 265098 |
+
"learning_rate": 7.929262820512822e-06,
|
| 265099 |
+
"loss": 0.4015,
|
| 265100 |
+
"step": 130265
|
| 265101 |
+
},
|
| 265102 |
+
{
|
| 265103 |
+
"epoch": 1041.79,
|
| 265104 |
+
"learning_rate": 7.929182692307693e-06,
|
| 265105 |
+
"loss": 0.7939,
|
| 265106 |
+
"step": 130270
|
| 265107 |
+
},
|
| 265108 |
+
{
|
| 265109 |
+
"epoch": 1041.83,
|
| 265110 |
+
"learning_rate": 7.929102564102564e-06,
|
| 265111 |
+
"loss": 0.7487,
|
| 265112 |
+
"step": 130275
|
| 265113 |
+
},
|
| 265114 |
+
{
|
| 265115 |
+
"epoch": 1041.87,
|
| 265116 |
+
"learning_rate": 7.929022435897437e-06,
|
| 265117 |
+
"loss": 0.3225,
|
| 265118 |
+
"step": 130280
|
| 265119 |
+
},
|
| 265120 |
+
{
|
| 265121 |
+
"epoch": 1041.91,
|
| 265122 |
+
"learning_rate": 7.928942307692309e-06,
|
| 265123 |
+
"loss": 0.2599,
|
| 265124 |
+
"step": 130285
|
| 265125 |
+
},
|
| 265126 |
+
{
|
| 265127 |
+
"epoch": 1041.95,
|
| 265128 |
+
"learning_rate": 7.92886217948718e-06,
|
| 265129 |
+
"loss": 0.6579,
|
| 265130 |
+
"step": 130290
|
| 265131 |
+
},
|
| 265132 |
+
{
|
| 265133 |
+
"epoch": 1041.99,
|
| 265134 |
+
"learning_rate": 7.928782051282053e-06,
|
| 265135 |
+
"loss": 0.8982,
|
| 265136 |
+
"step": 130295
|
| 265137 |
+
},
|
| 265138 |
+
{
|
| 265139 |
+
"epoch": 1042.0,
|
| 265140 |
+
"eval_loss": 0.4521143138408661,
|
| 265141 |
+
"eval_runtime": 36.4023,
|
| 265142 |
+
"eval_samples_per_second": 22.966,
|
| 265143 |
+
"eval_steps_per_second": 0.742,
|
| 265144 |
+
"eval_wer": 0.17621112989781665,
|
| 265145 |
+
"step": 130296
|
| 265146 |
+
},
|
| 265147 |
+
{
|
| 265148 |
+
"epoch": 1050.03,
|
| 265149 |
+
"learning_rate": 7.928701923076923e-06,
|
| 265150 |
+
"loss": 0.2942,
|
| 265151 |
+
"step": 130300
|
| 265152 |
+
},
|
| 265153 |
+
{
|
| 265154 |
+
"epoch": 1050.07,
|
| 265155 |
+
"learning_rate": 7.928621794871796e-06,
|
| 265156 |
+
"loss": 0.3089,
|
| 265157 |
+
"step": 130305
|
| 265158 |
+
},
|
| 265159 |
+
{
|
| 265160 |
+
"epoch": 1050.11,
|
| 265161 |
+
"learning_rate": 7.928541666666667e-06,
|
| 265162 |
+
"loss": 0.28,
|
| 265163 |
+
"step": 130310
|
| 265164 |
+
},
|
| 265165 |
+
{
|
| 265166 |
+
"epoch": 1050.15,
|
| 265167 |
+
"learning_rate": 7.928461538461539e-06,
|
| 265168 |
+
"loss": 0.3548,
|
| 265169 |
+
"step": 130315
|
| 265170 |
+
},
|
| 265171 |
+
{
|
| 265172 |
+
"epoch": 1050.19,
|
| 265173 |
+
"learning_rate": 7.92838141025641e-06,
|
| 265174 |
+
"loss": 0.8331,
|
| 265175 |
+
"step": 130320
|
| 265176 |
+
},
|
| 265177 |
+
{
|
| 265178 |
+
"epoch": 1050.23,
|
| 265179 |
+
"learning_rate": 7.928301282051283e-06,
|
| 265180 |
+
"loss": 0.6048,
|
| 265181 |
+
"step": 130325
|
| 265182 |
+
},
|
| 265183 |
+
{
|
| 265184 |
+
"epoch": 1050.27,
|
| 265185 |
+
"learning_rate": 7.928221153846154e-06,
|
| 265186 |
+
"loss": 0.2592,
|
| 265187 |
+
"step": 130330
|
| 265188 |
+
},
|
| 265189 |
+
{
|
| 265190 |
+
"epoch": 1050.31,
|
| 265191 |
+
"learning_rate": 7.928141025641026e-06,
|
| 265192 |
+
"loss": 0.282,
|
| 265193 |
+
"step": 130335
|
| 265194 |
+
},
|
| 265195 |
+
{
|
| 265196 |
+
"epoch": 1050.35,
|
| 265197 |
+
"learning_rate": 7.928060897435899e-06,
|
| 265198 |
+
"loss": 0.3886,
|
| 265199 |
+
"step": 130340
|
| 265200 |
+
},
|
| 265201 |
+
{
|
| 265202 |
+
"epoch": 1050.39,
|
| 265203 |
+
"learning_rate": 7.92798076923077e-06,
|
| 265204 |
+
"loss": 0.7989,
|
| 265205 |
+
"step": 130345
|
| 265206 |
+
},
|
| 265207 |
+
{
|
| 265208 |
+
"epoch": 1050.43,
|
| 265209 |
+
"learning_rate": 7.927900641025641e-06,
|
| 265210 |
+
"loss": 0.5287,
|
| 265211 |
+
"step": 130350
|
| 265212 |
+
},
|
| 265213 |
+
{
|
| 265214 |
+
"epoch": 1050.47,
|
| 265215 |
+
"learning_rate": 7.927820512820513e-06,
|
| 265216 |
+
"loss": 0.2833,
|
| 265217 |
+
"step": 130355
|
| 265218 |
+
},
|
| 265219 |
+
{
|
| 265220 |
+
"epoch": 1050.51,
|
| 265221 |
+
"learning_rate": 7.927740384615386e-06,
|
| 265222 |
+
"loss": 0.2952,
|
| 265223 |
+
"step": 130360
|
| 265224 |
+
},
|
| 265225 |
+
{
|
| 265226 |
+
"epoch": 1050.55,
|
| 265227 |
+
"learning_rate": 7.927660256410257e-06,
|
| 265228 |
+
"loss": 0.458,
|
| 265229 |
+
"step": 130365
|
| 265230 |
+
},
|
| 265231 |
+
{
|
| 265232 |
+
"epoch": 1050.59,
|
| 265233 |
+
"learning_rate": 7.927580128205129e-06,
|
| 265234 |
+
"loss": 0.8351,
|
| 265235 |
+
"step": 130370
|
| 265236 |
+
},
|
| 265237 |
+
{
|
| 265238 |
+
"epoch": 1050.63,
|
| 265239 |
+
"learning_rate": 7.9275e-06,
|
| 265240 |
+
"loss": 0.6399,
|
| 265241 |
+
"step": 130375
|
| 265242 |
+
},
|
| 265243 |
+
{
|
| 265244 |
+
"epoch": 1050.67,
|
| 265245 |
+
"learning_rate": 7.927419871794873e-06,
|
| 265246 |
+
"loss": 0.2709,
|
| 265247 |
+
"step": 130380
|
| 265248 |
+
},
|
| 265249 |
+
{
|
| 265250 |
+
"epoch": 1050.71,
|
| 265251 |
+
"learning_rate": 7.927339743589744e-06,
|
| 265252 |
+
"loss": 0.3261,
|
| 265253 |
+
"step": 130385
|
| 265254 |
+
},
|
| 265255 |
+
{
|
| 265256 |
+
"epoch": 1050.76,
|
| 265257 |
+
"learning_rate": 7.927259615384616e-06,
|
| 265258 |
+
"loss": 0.3979,
|
| 265259 |
+
"step": 130390
|
| 265260 |
+
},
|
| 265261 |
+
{
|
| 265262 |
+
"epoch": 1050.8,
|
| 265263 |
+
"learning_rate": 7.927179487179489e-06,
|
| 265264 |
+
"loss": 0.8701,
|
| 265265 |
+
"step": 130395
|
| 265266 |
+
},
|
| 265267 |
+
{
|
| 265268 |
+
"epoch": 1050.84,
|
| 265269 |
+
"learning_rate": 7.92709935897436e-06,
|
| 265270 |
+
"loss": 0.5942,
|
| 265271 |
+
"step": 130400
|
| 265272 |
+
},
|
| 265273 |
+
{
|
| 265274 |
+
"epoch": 1050.88,
|
| 265275 |
+
"learning_rate": 7.927019230769231e-06,
|
| 265276 |
+
"loss": 0.2871,
|
| 265277 |
+
"step": 130405
|
| 265278 |
+
},
|
| 265279 |
+
{
|
| 265280 |
+
"epoch": 1050.92,
|
| 265281 |
+
"learning_rate": 7.926939102564103e-06,
|
| 265282 |
+
"loss": 0.3117,
|
| 265283 |
+
"step": 130410
|
| 265284 |
+
},
|
| 265285 |
+
{
|
| 265286 |
+
"epoch": 1050.96,
|
| 265287 |
+
"learning_rate": 7.926858974358976e-06,
|
| 265288 |
+
"loss": 0.3888,
|
| 265289 |
+
"step": 130415
|
| 265290 |
+
},
|
| 265291 |
+
{
|
| 265292 |
+
"epoch": 1051.0,
|
| 265293 |
+
"learning_rate": 7.926778846153847e-06,
|
| 265294 |
+
"loss": 0.962,
|
| 265295 |
+
"step": 130420
|
| 265296 |
+
},
|
| 265297 |
+
{
|
| 265298 |
+
"epoch": 1051.0,
|
| 265299 |
+
"eval_loss": 0.36090824007987976,
|
| 265300 |
+
"eval_runtime": 36.3757,
|
| 265301 |
+
"eval_samples_per_second": 22.982,
|
| 265302 |
+
"eval_steps_per_second": 0.742,
|
| 265303 |
+
"eval_wer": 0.18075693360101122,
|
| 265304 |
+
"step": 130420
|
| 265305 |
}
|
| 265306 |
],
|
| 265307 |
+
"max_steps": 620000,
|
| 265308 |
"num_train_epochs": 5000,
|
| 265309 |
+
"total_flos": 3.670101011743939e+20,
|
| 265310 |
"trial_name": null,
|
| 265311 |
"trial_params": null
|
| 265312 |
}
|
model-bin/finetune/base/{checkpoint-129799 → checkpoint-130420}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630160919.4761047/events.out.tfevents.1630160919.86bb0ddabf9b.4092.191
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5f8725dfaf8783044ae15fe501347fdeffb8960b0f4a5cae50ddfb76b7ae4ac1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630161303.291062/events.out.tfevents.1630161303.86bb0ddabf9b.4092.193
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f36509ded478d023b3e6a15fb1f69e4e7a9c6ec408136a8435d648a5b0a0bf0c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630161692.6533394/events.out.tfevents.1630161692.86bb0ddabf9b.4092.195
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ac8785dacde214c4c38bf643881b198d190cd89457db86085d47747ba0087a35
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630162085.8766842/events.out.tfevents.1630162085.86bb0ddabf9b.4092.197
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bc714675afae783e7f4518cee99ffdd20ce5ead75255d96e0fb586f42d46167c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630162481.5400343/events.out.tfevents.1630162481.86bb0ddabf9b.4092.199
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b247231f4d868ae885320dc4af6e433dd766df680ac3373bfce081e6efb4d096
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630160919.86bb0ddabf9b.4092.190
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed53954dfb35bed2cf1abab6ae810faaf858af4ecea43081fdb05c2a0ae21273
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630161303.86bb0ddabf9b.4092.192
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ccc19d150895b3672edac61da04b15f1c0086182dc2877063608f78b127b630e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630161692.86bb0ddabf9b.4092.194
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3f63fa3db8125e0443461a1cedb49de0cfa487622fc504069ae3fb5b753443fe
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630162085.86bb0ddabf9b.4092.196
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f3853a0e0a244d413b733f1b90641b30d5d9a0141fc368e88ef5de60f47c4bbf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630162481.86bb0ddabf9b.4092.198
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f63c0c526cba80cf03271debf0009ff0ee3e2f2bb9b29a640245d865f77eabb
|
| 3 |
+
size 8622
|